>> No real world code is simple and elegant enough that you don't >> need to think through the complex invariants of the code

> Oh, please...

  sendDataWithTimeout(user_t *user, data_t *data, time_t timeout);

What happens if delay is 0?

Can user be NULL? What happens if it is NULL?

How large can 'data' be? Is there a limit? Will the send be chunked into multiple dispatches if 'data' is too large? Is that opaque to the caller? Does it matter to the caller?

If none of those questions are answered, the caller must delve into the sendDataWithTimeout() implementation to figure out the answers, and it is impossible to modify sendDataWithTimeout() without possibly breaking assumptions callers make based on what they've assumed from implementation of sendDataWithTimeout().

This applies equally well to blocks of code within a function that accept input and provide output. It helps to be able to reason about them as atomic units, without necessarily paying the code/maintenance cost of hoisting them into independent functions.

> This applies equally well to blocks of code within a function that accept input and provide output. It helps to be able to reason about them as atomic units, without necessarily paying the code/maintenance cost of hoisting them into independent functions.

This seems rather... crazy. Functions serve this purpose, avoiding creating function while still trying to think of your code block as a function is just absurd. Extract the function, if it deserves a comment and has parameters, and isn't an anonymous method for use with a higher order function, then make a real function out of instead of polluting the parent function with its implementation details.

The overhead of this additional function you're so worried about is nothing compared to the overhead of the mess that results by not extracting it. Your strategy favors long functions, yuk yuk yuk.

Listening to you talk, I'm horrified by the thought of what your code must look like; I bet your functions often require use of the scroll bar.

Functions for four lines of code is stupid when a simple one line comment would suffice.

No, it isn't; that's is precisely the purpose of a function and precisely the wrong way to comment.

Given the mental and work effort necessary to hoist out a function versus writing a simple comment, I think your position involves untenable trade-offs.

I'd much rather see code commenting the major steps of the function than a file riddled with tiny functions, all of which require breaking my train of thought, switching to a different section of the file to read, and then going back to the original function.

That's ludicrous compared to a one-liner "Send our token to the server" comment on a small block of code.

It's no effort, it's a simple automated refactoring, right click, extract method, choose name, done. I think your positions betrays your age and tells me you probably code in a text editor rather than a more powerful IDE. To fear extracting a method is just insane to me, a sign of past times.

Functions shouldn't have major steps, they should be one step each with a top level function calling the steps in sequence. A function should do one conceptual thing, or be a delegator to sequence other functions.

Mixing levels of abstraction where all the steps are inlined into the larger master function that also sequences them is IMHO a code smell that demands fixing. Kent Beck had quite a bit to say about this very thing in Smalltalk Best Practice Patterns.

Like I said, I bet you use the scroll bar a lot.

> Like I said, I bet you use the scroll bar a lot.

Says the fellow that hoists incredibly simple code out to independent functions.

On the subject of age, since you broached it -- Given your seriously arguing for not commenting, coupled with the naive hubris so often associated with youth, I can only assume that you're ether young, inexperienced, and/or do little work in maintaining production code across the years.

Wrong on all counts. I'm old, experienced, and maintain decades old systems.

And I'm not arguing for not commenting, I'm arguing for writing more expressive code so that function names are the comments and additional comments are not required. Nothing beats executable comments.

So, if the method was commented, would you:

(A) assume the comments are correct and skip reading what the code does?

(B) use comments as guidance and read code anyway?

The first seems awfully risky to me and poor advice to any programmer trying to write solid code.

> (A) assume the comments are correct and skip reading what the code does?

Yes, unless I was working on that particular piece of code. There's often far too much code in a code base to waste time reading it all, and reading the comments can provide quick understanding.

> The first seems awfully risky to me and poor advice to any programmer trying to write solid code.

Why? Bad code is bad code. Writing bad comments (or no comments) means you're writing bad code. Don't write bad code, and then write documentation to help yourself think through why it's not bad code, and tests to both help you think through it, and verify that it's not bad.